OMRAS 2 Use-case analysis

We can look at this problem in several ways. One is to see what others have done, then refine and extend it. The advantage of this method is that one is fairly likely (?) to cover the most important/popular use cases. Another would be to make a principled attempt to cover all possible use cases. The obvious disadvantage of this is the amount of effort/time required.

A 'middle way' might be to review the literature with a view to finding out what the few most significant 'parameters of use' seem to be, and to use them in a more or less systematic way. This can be supplemented by asking potential users. At present, this seems the likely course for OMRAS 2.

In looking at the MIREX tasks set in 2005 & 2006, it's clear that they are strongly oriented towards low-level, rather than system-level tasks. In fact, even in the main ISMIR papers, there is very little literature mentioning system-level uses, tasks and/or evaluation. So in some respects we will be pioneers.

OMRAS 2 Use-case analysis

Infrastructural levels of tasks and MIR programs

	Level
S	System-level	Integrates multiple coordinated program-modules (including query-interface and results-display), working seamlessly and transparently to present a complete picture of search results without further processing or user-interaction
H	High-level	Integrates and coordinates the interaction of multiple modules in order to pursue a very specific analytical or search task; may involve manual processing of queries and/or results
L	Low-level	A task, normally involving a single module, aiming to answer a specific research question; no integration of user-interface for queries or result-display

Examples of how these categories might be applied to MIREX tasks

MIREX 2006

	Level
Audio Beat Tracking	L
Audio Melody Extraction	L/H
Audio Music Similarity and Retrieval	H/S
Audio Cover Song	S
Audio Onset Detection	L
Audio Tempo Extraction	L
QBSH: Query by Singing/Humming	H/S
Score Following	H
Symbolic Melodic Similarity	H/S

MIREX 2005

	Level
Audio Artist Identification	H
Audio Drum Detection	L
Audio Genre Classification	H
Audio Melody Extraction	H
Audio Onset Detection	L
Audio Tempo Extraction	L
Audio and Symbolic Key Finding	H
Symbolic Genre Classification	H
Symbolic Melodic Similarity	H

An extended similarity scale for content-based MIR (based on Don Byrd's)

A Detailed audio characteristics in common

1	Same	Different	Comments & Examples
	Music		Audio fingerprinting; Same digital encoding - just noisy query
	Arrangement
	Venue
	Artists
	Session
	Performance
	Recording

2	Same	Different	Comments & Examples
	Music	Recording	Remixes; Grateful Dead audience recordings; early cylinder 'mass-produced' recordings (there are surviving examples where several recording machines were set going simultaneously; NB this leads - accidentally - to stereo recordings!)
	Arrangement
	Venue
	Artists
	Session
	Performance

3	Same	Different	Comments & Examples
	Music	Performance	Immediate studio retakes; e.g., Miles Davis Kind of Blue CD (Flamenco Sketches)
	Arrangement	Recording
	Venue
	Artists
	Session

4	Same	Different	Comments & Examples
	Music	Session	'Reconstructive' retakes' - e.g. opera or symphony recordings over more than one day
	Arrangement	Performance
	Venue	Recording
	Artists

5	Same	Different	Comments & Examples
	Music	Artists	'Remedial' retakes' - e.g. overdubbing a different singer
	Arrangement	Session
	Venue	Performance
		Recording

B Few or no detailed audio characteristics in common

4a	Same	Different	Comments & Examples
	Music	Venue	Concert recordings
	Arrangement	Session
	Artists	Performance
		Recording

5	Same	Different	Comments & Examples
	Music	Artists
	Arrangement	Venue
		Session
		Performance
		Recording

6	Same	Different	Comments & Examples
	Music	Arrangement
		Artists
		Venue
		Session
		Performance
		Recording

7	Same	Different	Comments & Examples
		Music
		Arrangement
		Artists
		Venue
		Session
		Performance
		Recording

-- TimCrawford - 23 Jan 2007

Main

Webs
Main
Sandbox
TWiki

Copyright © by the contributing authors. All material on this collaboration platform is the property of the contributing authors.
Ideas, requests, problems regarding OMRAS2? Send feedback